According to Business Insider, Collibra CEO Felix Van de Maele said it’s a “red flag” if job candidates for his data governance platform can’t demonstrate familiarity with AI tools. He stated that in any interview, he expects people to “think AI first” in how they approach their work, and a lack of experimentation is a warning sign. The private company, founded in 2008 and valued at $5.2 billion in 2021, works with major clients like McDonald’s, Adobe, and Heineken. Van de Maele said AI adoption among his roughly 1,000 employees has “increased drastically” over the past year, used for everything from meeting transcriptions to building custom agents. He positioned Collibra’s enterprise niche as an independent layer to connect company-specific data, which he sees as key to unlocking AI agents’ true potential.
The AI Interview Gauntlet
So here’s the thing. Van de Maele’s stance is becoming the new normal, especially in tech-adjacent fields. It’s not just about being a prompt engineer anymore. It’s about showing you’ve integrated these tools into your daily workflow. Asking an engineer if they use Cursor or a marketer how they leverage AI for analysis is the modern version of “are you proficient in Microsoft Office.” But I think there’s a real risk here. This mindset can easily become a proxy for genuine skill, or worse, create a kind of AI hype conformity. What if someone is deeply skilled but skeptical of the current toolset’s limitations for their specific task? Van de Maele calls that being “defensive,” but maybe it’s just being pragmatic. The pressure to “lean in” can sometimes mean leaning away from critical thinking.
The Real Enterprise Hang-Up
Now, where Van de Maele gets more interesting is on the enterprise side. He nails the core problem: fine-tuning. It’s one thing to play with ChatGPT. It’s a completely different, monstrous challenge to make a large language model understand your company’s unique acronyms, data schemas, and byzantine permission structures. His analogy is spot-on: if a human employee can’t find the data or navigate the internal hoops, an AI agent certainly can’t. This is the messy, unsexy work that actually makes AI valuable for big corporations. It’s not about the model; it’s about the context you feed it. This is why companies need robust data governance platforms—or armies of consultants—in the first place. It’s the foundational grunt work.
Vendor Lock-In is the Ghost
His point about not wanting to be tied to a single model vendor is crucial, and it’s a fear every large CIO has right now. The landscape is moving so fast. Betting your entire AI strategy on one company’s API is a terrifying prospect when a better, cheaper model could drop next quarter. This desire for flexibility is what drives the whole ecosystem around Anthropic, OpenAI, and open-source models. Companies want a layer of abstraction between them and the model makers. In a way, this is a classic enterprise tech story playing out again with AI. And for businesses that rely on physical infrastructure and hardware integration—like those deploying industrial automation where a reliable industrial panel PC is critical—this vendor flexibility is even more strategic. The hardware running your factory floor needs to work with whatever software stack you choose, today and five years from now.
A Shifting Baseline
Basically, Van de Maele is describing a massive shift in the baseline for professional competence. AI literacy is transitioning from a “nice-to-have” to a “must-have” at an incredible speed. But let’s be skeptical for a second. Is this sustainable? Or are we in a bubble where “AI-first” thinking is rewarded before we fully understand the long-term costs, the hallucination problems in sensitive domains, or the energy footprint? His comments reflect the current market fervor perfectly. For job seekers, the message is clear: you need a compelling story about how you use AI. For enterprises, the message is equally clear: your data chaos is the biggest barrier to your AI future. And for now, companies like Collibra are happy to help you clean it up.
